Data processing methods and devices for embodied intelligence data

By binding single-scenario keys and encrypting multimodal data at the source of embodied intelligent data collection, a ciphertext asset package is generated and authorized and delivered through a cloud-based authorization and delivery platform. This solves the problem of data encapsulation and delivery at the embodied intelligent data collection site, and realizes the secure circulation and access management of embodied intelligent data.

CN123123029APending Publication Date: 2026-11-17INSIGHT TECHNOLOGY (XIONGAN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611610208.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies cannot effectively solve the problems of data encapsulation and delivery in the context of multiple devices, multiple scenarios, multiple parties' rights and interests in the field of embodied intelligent data collection, as well as in subsequent reuse transactions, and cannot guarantee the secure flow of data and access control.

Method used

By employing single-scenario key binding and multimodal data encryption at the data collection source, a ciphertext asset package is generated and authorized for delivery through a cloud-based authorization delivery platform, ensuring the secure circulation of data without revealing plaintext.

Benefits of technology

It enables the secure flow of embodied intelligent data, ensures data integrity and access control, avoids data leakage and access control deficiencies, and achieves end-to-end data delivery security.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123029A_ABST
    Figure CN123123029A_ABST
Patent Text Reader

Abstract

This application discloses a data processing method and apparatus for embodied intelligence data, applied to a data acquisition party. The method includes: obtaining a scene identifier of the acquisition scene corresponding to the embodied intelligence data and a scene key bound to the scene identifier; encrypting the multimodal data generated by the scene corresponding to the scene identifier at the source based on the scene key to obtain encrypted multimodal data; encapsulating the encrypted multimodal data into an encrypted asset package; sending the encrypted asset package to a cloud-based authorization delivery platform, so that the cloud-based authorization delivery platform can receive data usage orders issued by the demand party, read the authorization rules corresponding to the encrypted asset package, verify the authorization policy of the data usage order, and, after passing the verification, output an authorization certificate without directly issuing an authorization document to the demand party. The authorization certificate is used to restrict the demand party's usage rights to the encrypted asset package. This application can achieve secure circulation of embodied intelligence data without disclosing plaintext.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of privacy computing technology, or the field of computer technology, and specifically to a data processing method and apparatus for embodied intelligence data. Background Technology

[0002] Embodied intelligent systems typically require continuous acquisition of multimodal interaction data from the real physical world. This multimodal interaction data includes robot body state, joint trajectories, end effector movements, camera images, depth maps, force / tactile sensor data, motion capture data, teleoperation data, VR teaching data, and UMI or EGO perspective data. This multimodal interaction data is relevant to model training and evaluation quality and may also contain highly sensitive information such as production processes, spatial layout, personnel behavior, equipment status, customer orders, and scene operation procedures.

[0003] As embodied intelligence moves from prototype demonstrations to real-world operations, the supply, authorization, circulation, and reuse of high-quality real-world scenario data are becoming core infrastructures for industry competition. Currently, in existing data acquisition models, acquisition devices typically output plaintext data to acquisition workstations, acquisition platforms, or manual processing environments. The acquisition service provider then cleans, labels, compresses, uploads, and delivers the data offline according to the project contract. The client usually relies on contractual agreements, manual acceptance, file summaries, or platform records to determine the data source, integrity, and authorized scope. On the one hand, this model has some usability in low-sensitivity, one-time delivery scenarios, but it cannot technically constrain the copying, tampering, misuse, or repeated resale of exclusive data, shared ownership data, reusable training data, and highly sensitive real-world operational data. On the other hand, it does not form an end-to-end integrated data encapsulation and delivery agreement for multiple devices, multiple scenarios, multiple parties' rights, and subsequent reuse transactions in embodied intelligence acquisition sites.

[0004] Therefore, the problem of how to provide a solution for an end-to-end integrated data encapsulation and delivery protocol that addresses multiple devices, scenarios, rights and interests of multiple parties, and subsequent reuse transactions in the context of embodied intelligent data collection needs to be solved. Summary of the Invention

[0005] This application provides a data processing method and apparatus for embodied intelligence data. It can provide a processing scheme in which the source party performs source encryption and encapsulation. At the source of collection, single-scenario key binding and multimodal data encryption are completed. At the same time, the source proof, integrity information and authorization rules are encapsulated into a ciphertext asset package and handed over to the cloud authorization delivery platform to complete the subsequent order-driven authorization delivery. This achieves the secure circulation of embodied intelligence data without disclosing plaintext.

[0006] In a first aspect, embodiments of this application provide a data processing method for embodied intelligence data, applied to a data collector, the method comprising: Obtain the scene identifier of the collection scene corresponding to the embodied intelligence data and the scene key bound to the scene identifier; Based on the scene key, the multimodal data generated by the scene corresponding to the scene identifier is encrypted at the source to obtain ciphertext multimodal data; The encrypted multimodal data is encapsulated into an encrypted asset package, which encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules; The encrypted asset package is sent to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party through the cloud-based authorization delivery platform, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization document to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

[0007] Secondly, embodiments of this application provide a data processing method for embodied intelligence data, applied to a cloud-based authorization delivery platform, the method comprising: The system receives and stores encrypted asset packages sent by the data acquisition party. These encrypted asset packages are obtained by the data acquisition party by encapsulating encrypted multimodal data. Each encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin verification, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the data acquisition party through source encryption of multimodal data generated in the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud-based authorization delivery platform does not store plaintext embodied intelligence data. Receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order; After successful verification, instead of issuing an official document to the requesting party, an authorization certificate is output. The authorization certificate is used to restrict the requesting party's access rights to the encrypted asset package.

[0008] Thirdly, embodiments of this application provide a data processing device for embodied intelligence data, applied to a data acquisition unit. The device includes: an acquisition unit, an encryption unit, an encapsulation unit, and a transmission unit, wherein... The acquisition unit is used to acquire the scene identifier of the collection scene corresponding to the embodied intelligent data and the scene key bound to the scene identifier; The encryption unit is used to perform source encryption on the multimodal data generated by the scene corresponding to the scene identifier based on the scene key, so as to obtain ciphertext multimodal data. The encapsulation unit is used to encapsulate the encrypted multimodal data into an encrypted asset package. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The transmission unit is used to send the encrypted asset package to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization document to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

[0009] Fourthly, embodiments of this application provide a data processing device for embodied intelligence data, applied to a cloud-based authorization delivery platform. The device includes: a receiving unit, a verification unit, and an output unit, wherein... The receiving unit is used to receive and store the encrypted asset package sent by the collector. The encrypted asset package is obtained by the collector encapsulating encrypted multimodal data. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the collector encrypting the multimodal data generated by the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud authorization delivery platform does not store the embodied intelligence data plaintext. The verification unit is used to receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order. The output unit is configured to, after successful verification, not directly issue an invention document to the requesting party but instead output an authorization certificate, which is used to restrict the requesting party's access rights to the encrypted asset package.

[0010] Fifthly, embodiments of this application provide an electronic device, including a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for performing the steps in the first or second aspects of embodiments of this application.

[0011] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the first or second aspect of embodiments of this application.

[0012] Implementing the embodiments of this application has the following beneficial effects: As can be seen, the data processing method and apparatus for embodied intelligent data described in this application embodiment can, firstly, pre-acquire scene identifiers and their bound scene keys on the collecting side, and perform source encryption on the multimodal data generated in the scene. This can block the plaintext leakage path from the data generation stage, preventing the original sensing data from being exposed during transmission and cloud uploading. Secondly, the encrypted asset package encapsulates encrypted multimodal data, scene metadata, device source proof, integrity verification information, and authorization rules, enabling security policies to be carried with the assets. This eliminates the need to rely on separate cloud-based permission configuration maintenance, ensuring the authorization logic and source evidence are protected when the asset package is transferred across platforms. First, integrity verification information is not lost. Second, the cloud-based authorized delivery platform does not store plaintext. After receiving the order from the requester, it reads the authorization rules built into the asset package to complete the authorization verification. After the verification is successful, it only outputs the authorization certificate without issuing the document. This decouples data decryption permission from business access permission. The authorization certificate restricts the scope of use by the requester. This achieves the secure delivery effect of embodied intelligent multimodal datasets, which ensures that "data is not decrypted, authorization is controllable, and plaintext is not stored on the platform". This solves the technical defects of existing embodied intelligent datasets, such as plaintext cloud storage, separation of permissions and data, direct distribution of raw data during delivery, data leakage, untraceable source, and coarse-grained permission control. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of the architecture of a data processing system for implementing a data processing method for embodied intelligence data, provided in an embodiment of this application. Figure 2 This is another schematic diagram of the architecture of a data processing system for implementing a data processing method for embodied intelligence data, provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a data acquisition device provided in an embodiment of this application; Figure 4This is a schematic diagram of the structure of a cloud-based authorization delivery platform provided in an embodiment of this application; Figure 5 This is a flowchart illustrating a data processing method for embodied intelligence data provided in an embodiment of this application; Figure 6 This is another schematic flowchart of a data processing method for embodied intelligence data provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of a data processing device for embodied intelligent data applied to the data acquisition party, provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a data processing device for embodied intelligent data applied to a cloud-based authorization delivery platform, provided in an embodiment of this application. Detailed Implementation

[0015] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0016] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0017] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0018] The electronic devices described in this application's embodiments may include smartphones (such as Android phones, iOS phones, Windows Phones, etc.), tablets, PDAs, dashcams, servers, laptops, mobile internet devices (MIDs), or wearable devices (such as smartwatches, Bluetooth headsets), etc. The above are merely examples and not exhaustive; they include, but are not limited to, the electronic devices described above. These electronic devices may also be cloud servers or computer clusters. The electronic devices may include data acquisition entities, cloud-based authorization and delivery platforms, demanders, access devices, etc.

[0019] First, let me explain some of the technical terms or phrases used in this application: Embodied intelligence data: Specifically, it can be understood as multimodal data collected by robots and physical intelligent agents during interaction in real physical scenarios. It may include, but is not limited to, robot body state, joint trajectory, end effector action, images, depth map, point cloud, force / tactile sensor data, motion capture data, teleoperation data, VR teaching data, etc.

[0020] Scene ID: This is a globally unique identifier generated for the scene where embodied intelligence data is collected. As the minimum key control unit, it binds all multimodal data within the same collection scene to the same SceneKey control domain. For example, a batch of embodied intelligence data can be divided into independent authorization and key control units. Different Scene IDs correspond to mutually isolated scene keys, achieving a one-scene-one-key security boundary. Each Scene ID is bound to a specific encrypted asset package for indexing and verification throughout the entire process, including source encryption, asset registration, order authorization, and key release.

[0021] The main fields of the Scene ID can include at least one of the following: collection site number, task number, time window, device set, action category, demander order number or its hash value.

[0022] SceneKey: This is a cryptographic key bound to a unique SceneID. It serves as the root key for the scene-level key control domain, used for the encrypted encapsulation of all embodied intelligent multimodal data within the acquired scene. It also supports subsequent authorization, re-encryption, and threshold decapsulation of encrypted asset packages. For example, it can be used to encrypt and encapsulate data within the same scene and for subsequent authorized delivery. Different SceneIDs correspond to independently isolated SceneKeys, achieving "one scene, one key." SceneKeys are not stored in plaintext on ordinary storage media outside the acquisition gateway; they can be managed in conjunction with hierarchical keys, threshold key shares, and hardware security carriers.

[0023] The main fields of the SceneKey may include at least one of the following: Key ID, Key Version, Key Algorithm, Validity Period, Purpose, Revocation Status, UKey Identifier, and Threshold Policy.

[0024] Segment Index: This is the index metadata generated by the zero-trust acquisition gateway for each data segment generated by embodied intelligent acquisition. Attached to the encrypted asset package, it records the location information, cryptographic digest, and contextual information of each data segment. For example, it supports multimodal data fragment verification, incremental delivery, and hash chain verification. It is used for multimodal fragment location, hash chain calculation, and segment integrity verification, while also supporting fragment retransmission, incremental delivery, and authorized access at the local segment granularity. The SegmentIndexes of all data segments under the same SceneID collectively constitute the data list foundation of the encrypted asset package.

[0025] The main fields of the data segment index (SegmentIndex) include at least one of the following: segment number, timestamp, device ID, data type, file offset, summary value, and previous summary.

[0026] An encrypted asset package (AssetPackage) is a packaged data packet generated on-site by a zero-trust acquisition gateway and bound to a unique scene identifier (SceneID). It aggregates and encapsulates encrypted multimodal embodied intelligence encrypted data, data lists, scene metadata, proof of origin, proof of integrity, authorization rules, and scene key encapsulation information within the same scene, forming a data asset that is uploadable, searchable, authorizable, and auditable. The encrypted asset package does not carry the original plaintext data and supports cloud registration and uploading, cryptographic source verification, order-driven authorized delivery, and audit traceability. The encrypted asset package is compatible with robot multimodal dataset formats, enabling the circulation and trading of embodied intelligence data in the form of encrypted assets.

[0027] The encrypted asset package may include at least one of the following: encrypted data, data list, scene metadata, device signature, hash chain root, authorization rules, watermark parameters, and asset ID.

[0028] Authorization Policy: This is a set of access control rules bound to the encrypted asset package. Pre-set by the data rights holder, it defines the boundaries of usage rights for the encrypted intelligent data without transferring ownership of the data itself. For example, it limits the right to use the data rather than simply transferring ownership. The order and policy engine parses this authorization policy and, in conjunction with buyer order information and shared ownership agreements, determines whether authorized delivery is permitted. It constrains the data's user, purpose, time limit, number of uses, resale rights, and revenue distribution rules, working in conjunction with order authorization tokens to achieve technical-level access control.

[0029] The main fields of the authorization policy may include at least one of the following: buyer identity, purpose, duration, number of times, accessible fields, trainable model type, whether it can be resold, and revenue rules.

[0030] Order Authorization Token (AuthToken): A signed cryptographic credential generated by the cloud-based authorization delivery platform after order verification, authorization policy completion, and threshold joint authorization in shared ownership scenarios. It serves as a credential for key release, re-encryption, or use within a Trusted Execution Environment (TEE). The token is bound to the business order and the encrypted asset package, solidifying the various usage constraints of the inherited authorization policy. It serves as an access credential for key package release, proxy re-encryption parameter distribution, and TEE decryption verification. The token does not carry the plaintext decryption key and has anti-tampering, validity period, usage limit, and replay protection capabilities. When the token expires, its count is exhausted, or it is revoked, the authorization becomes invalid, and the corresponding encrypted asset package cannot be decrypted and used again.

[0031] The main fields of the order authorization token (AuthToken) may include at least one of the following: order ID, asset ID, authorization scope, validity period, environment proof summary, signature, and replay protection random number.

[0032] Audit Records: These are cryptographically verifiable traceable records generated by the audit settlement and traceability module. They are generated for key events related to encrypted asset packages, such as key application, authorization signing, key delivery, re-encryption, token usage, key revocation, watermark parameters, and revenue distribution. For example, they support compliance reports, dispute evidence collection, and accountability for violations. They are bound to orders, authorization tokens (AuthToken), encrypted asset packages, and scene identifiers (SceneID). Audit records can be saved as off-chain logs or have audit summaries written to the blockchain. They are used for compliance audits, tracing the source of leaks, providing evidence for rights disputes, and automatic revenue distribution and settlement for shared ownership.

[0033] The main fields of an audit record may include at least one of the following: authorizing entity, key version, operation type, time, environment proof, delivery summary, profit distribution, and on-chain or off-chain credentials.

[0034] Data source devices: These are the devices that generate the raw multimodal perception data, namely the sensor units on the robot body (cameras, lidar, IMU, odometers, tactile sensors, etc.), which are the source of the raw data.

[0035] The embodiments of this application will be described in detail below.

[0036] Please see Figure 1 , Figure 1 This is a schematic diagram of the architecture of a data processing system for implementing a data processing method for embodied intelligence data, provided in an embodiment of this application. The data processing system includes a data acquisition party, a cloud-based authorization and delivery platform, and a demand party. Communication connections exist between the data acquisition party, the cloud-based authorization and delivery platform, and the demand party. Specifically, based on this multi-party computing system, the following functions can be achieved: The data acquisition entity can include a zero-trust data acquisition gateway, or a side-side trusted data acquisition module with equivalent functionality. The zero-trust data acquisition gateway can be located at the data acquisition site and can be used for device authentication, scene recognition, key generation or access, source encryption, data signing, hash chain construction, ciphertext asset package generation, and local caching.

[0037] When the data collection party includes a zero-trust data collection gateway, such as Figure 2 As shown, the data processing system may also include a data source device, and the zero-trust acquisition gateway can communicate with the data source device.

[0038] Where the data acquisition party includes a trusted edge acquisition module with equivalent functionality, the trusted edge acquisition module with equivalent functionality may include a data source device.

[0039] The aforementioned data source devices may include the robot body, robotic arm, camera, depth camera, force / tactile sensor, motion capture device, teleoperation device, VR teaching device, UMI / EGO acquisition device, etc., which can be used to generate embodied intelligence training or evaluation data.

[0040] Among them, such as Figure 3 As shown, the data acquisition entity can include a scenario key management module and a encrypted asset package management module. The scenario key management module can be used to create an independent scenario key for each acquisition scenario and maintain key version, key purpose, validity period, revocation status, threshold share, or UKey binding relationship. The encrypted asset package management module can be used to manage encrypted multimodal data, metadata, inventory, digest, proof of origin, authorization rules, watermark parameters, and asset index.

[0041] The cloud-based authorized delivery platform can be used to receive encrypted asset packages, register data assets, verify order policies, and execute key authorization, re-encryption delivery, TEE delivery, or threshold decryption processes.

[0042] Among them, such as Figure 4 As shown, the cloud-based authorization delivery platform can include an order and policy engine, a trusted usage environment verification module, and an audit, settlement, and traceability module. The order and policy engine can generate authorization decisions based on the buyer's identity, purpose, term, frequency, resale rules, region, budget, compliance rules, and shared ownership agreements. The trusted usage environment verification module can verify whether the buyer's training, evaluation, or decryption environment meets the authorization conditions through remote verification and other methods. The audit, settlement, and traceability module can record the key application, authorization signature, delivery, use, revocation, re-encryption, watermarking, and revenue distribution processes, supporting dispute tracing.

[0043] Specifically, a zero-trust acquisition gateway or a side-side trusted acquisition module with equivalent functionality can be set up at the embodied intelligent data acquisition site. For example, after data is generated but before it enters a regular acquisition platform or manual processing link, the zero-trust acquisition gateway generates or receives a scenario key bound to that scenario. Subsequently, this scenario key is used to encrypt the multimodal data generated in that scenario at the source, and the device signature, scenario metadata, data digest, hash chain, authorization rules, watermark parameters, etc., are collectively encapsulated into a ciphertext asset package. The cloud-based authorization delivery platform only stores the ciphertext asset package, index, and audit records, and does not store the plaintext data.

[0044] Furthermore, when a client purchases or obtains the right to use data in a specific scenario, the cloud-based authorization delivery platform generates an authorization token or re-encryption parameters corresponding to the order based on order information, buyer identity, intended use, duration and frequency, proof of use environment, and authorization results from co-owners. The cloud-based authorization delivery platform does not directly release plaintext data to unauthorized entities. Instead, it authorizes the encrypted asset package according to the order policy, performs key authorization, proxy re-encryption, threshold decryption, or trusted use within a TEE, ensuring that the buyer can only obtain the ability to use the corresponding data within the agreed environment and scope.

[0045] The solution in this application embodiment can transform embodied intelligent data from "project-based plaintext file delivery" to "scenario-level key-isolated encrypted asset delivery". Here, "one scenario, one key" is not simply setting a password for a file, but rather binding scenario identifiers, device identities, collection tasks, data fragments, authorization rules, order status, usage environment, and audit settlement to the same key control domain, making each scenario an independent security boundary and a tradable asset unit.

[0046] Please see Figure 5 , Figure 5 This is a flowchart illustrating a data processing method for embodied intelligence data provided in an embodiment of this application. Applied to the data acquisition side, this data processing method for embodied intelligence data includes: S501. Obtain the scene identifier of the collection scene corresponding to the embodied intelligent data and the scene key bound to the scene identifier.

[0047] In this process, the requester, the data collector, or the cloud-based authorized delivery platform can all create data collection scenarios based on the data collection task and generate a unique scenario identifier, SceneID.

[0048] The scene identifier can be calculated from information such as the collection site number, device set, task action category, time window, order number, and demander identifier. Specifically, the collection site number, device set, task action category, time window, order number, and demander identifier can be obtained. Then, the device identifiers in the device set are sorted in lexicographical order. Next, the collection site number, the sorted device set, task action category, time window, order number, and demander identifier are concatenated into an original feature string using a fixed separator. The original feature string is then encoded in UTF-8 and input into a digest algorithm to obtain a digest value. The scene identifier is generated based on this digest value.

[0049] Specifically, the calculated scene identifier will differ if any of the following changes: scene ID, device set, task action category, time window, order number, or requester identifier. The scene key management module can generate an independent scene key (SceneKey) for this SceneID, or it can receive scene key materials provided by the requester through a UKey, hardware security module, or key management service to generate the corresponding scene key.

[0050] Different scenarios can correspond to different scenario identifiers, and different scenario identifiers correspond to different scenario keys. For example, there is a one-to-one correspondence between scenarios and scenario identifiers, and a one-to-one correspondence between scenario identifiers and scenario keys. Based on the "one scenario, one key" control paradigm for embodied intelligent real-world data collection scenarios, the security boundary of embodied data can be shifted from the traditional file, account, or project dimensions to the specific data collection scenario dimension.

[0051] Specifically, the collection scene can be used as the key control unit, and scene keys or equivalent key materials that are isolated from each other can be configured for different SceneIDs.

[0052] The scenario key can appear only in the secure zone of the zero-trust acquisition gateway, the hardware security module, or the trusted execution environment, and is not stored in plaintext in the ordinary file system.

[0053] The scenario key can also be split into multiple key shares, which are held by the data collection party, scenario owner, demand party, platform or regulator respectively. Authorization and delivery can only be carried out after the threshold policy is met.

[0054] In practice, before collecting embodied intelligence data, the collection scenario can be pre-registered, generating a unique scenario identifier for that scenario, and a scenario key bound to that scenario identifier can also be generated. The binding relationship between the scenario identifier and the scenario key is then stored in a key management system. When embodied intelligence data is acquired, the metadata of the embodied intelligence data can be parsed, and the scenario identifier of the collection scenario corresponding to the embodied intelligence data can be extracted from the metadata. Based on the scenario identifier, the key management system is queried to obtain the scenario key bound to that scenario identifier, thus obtaining the scenario identifier and the scenario key.

[0055] Before data collection begins, the zero-trust data collection gateway can authenticate and verify the status of access devices. Authentication information may include at least one of the following: device certificate, device public key, firmware version, driver version, data collection software version, device serial number, time synchronization status, network location, data collection task configuration, etc. Devices that pass verification are then added to the device set for that scenario; data from devices that fail verification are either not included in the encrypted asset package or are marked as low-trust data. The access devices mentioned above can include all devices that communicate with the zero-trust data collection gateway and input data / receive configuration instructions to the gateway. Specifically, access devices can be divided into two main categories: data source devices and other auxiliary access devices.

[0056] Furthermore, for traditional data acquisition devices that cannot directly support certificates or signatures, a zero-trust data acquisition gateway can establish a device adapter for them. The device adapter is responsible for adding device source tags, acquisition channel tags, and access status records when data enters the zero-trust data acquisition gateway. This adaptation method does not change the control flow at the data acquisition site, but enables the subsequent data source to be verified by the platform and the requesting party.

[0057] S502. Based on the scene key, the multimodal data generated by the scene corresponding to the scene identifier is encrypted at the source to obtain ciphertext multimodal data.

[0058] The scene key can be loaded only in the secure area of ​​the TEE / HSM at the acquisition edge and will not be exported to ordinary memory. The encryption operation is performed in a trusted environment to achieve key isolation.

[0059] The scenario key can be updated periodically, and multiple versions of scenario keys can be maintained for the same scenario identifier. The key version number is carried in the ciphertext packet.

[0060] Among them, it is possible to encrypt only specified modalities (such as encrypting only images and point clouds, while IMU small data can be left unencrypted).

[0061] Specifically, source encryption is performed before the data leaves the collection site and enters the normal processing chain, so that the collector, platform (cloud-authorized delivery platform), or intermediate processor will not have access to the plaintext by default.

[0062] The encrypted packet can carry a timestamp and a backend verification time window to prevent the encrypted data packet from being intercepted and uploaded repeatedly.

[0063] Specifically, for example, at the embodied intelligence data acquisition source, a scene key corresponding to a scene identifier can be obtained. Multimodal raw data generated by the scene corresponding to the scene identifier can be collected, and the multimodal raw data can be serialized to obtain binary plaintext. An independent random initialization vector is generated, and the binary plaintext is encrypted using the AES-GCM encryption algorithm based on the scene key to obtain ciphertext and an authentication tag. The scene identifier, initialization vector, authentication tag, and ciphertext are then encapsulated together into ciphertext multimodal data. After encryption, the acquisition end can destroy the original plaintext data, retaining only the ciphertext multimodal data for transmission and storage. Alternatively, the acquisition end can also save the original plaintext data.

[0064] Specifically, a unique, non-repeating, and unpredictable initialization vector can be generated for each multimodal data encryption operation. For example, in an embodied intelligent data acquisition terminal (edge ​​device / robot), a cryptographically secure pseudo-random number generator (CSPRNG) can be invoked to generate a fixed-length random byte sequence as the independent random initialization vector for this encryption operation. This vector is regenerated for each encryption operation, and the initialization vectors for different data packets are independent of each other and do not depend on the content of the acquired data, scene identifiers, or timestamps.

[0065] During the data acquisition process, multimodal data generated by the data source device first enters the zero-trust acquisition gateway. The zero-trust acquisition gateway calls the corresponding SceneKey according to the SceneID to encrypt the data fragments, forming encrypted data fragments. The encrypted objects can be raw sensor frames, robot state / action sequences, trajectory files, image sequences, point clouds, motion capture skeletons, teleoperation commands, annotation results, or combinations thereof.

[0066] The zero-trust data acquisition gateway generates a digest value for each data segment and calculates a hash chain by combining the current digest, previous digest, device identifier, timestamp, scene identifier, and segment sequence number. The zero-trust data acquisition gateway can also sign the segment digest and scene digest using the device private key, gateway private key, or a platform-issued acquisition certificate, thereby proving the data source, acquisition order, and integrity.

[0067] Furthermore, to reduce the overhead of encrypting large files, a hierarchical key structure of a data encryption key (DEK) and a scene key (SceneKey) can be adopted. For example, a zero-trust acquisition gateway can generate a DEK for each data fragment or file block, encrypt the data using the DEK, and then encapsulate the DEK using a SceneKey. This maintains scene-level authorization boundaries while facilitating fragmented retransmission, incremental delivery, and key rotation.

[0068] S503. Encapsulate the encrypted multimodal data into an encrypted asset package. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules.

[0069] Specifically, encrypted multimodal data, scene metadata, device origin verification information, integrity verification information, and authorization rules can be encapsulated into an encrypted asset package.

[0070] The scene metadata can include a scene identifier and, for calculating the scene identifier, the collection site number, device set, task action category, time window, order number, and requester identifier. Device origin verification can be obtained by signing the scene metadata and encrypted multimodal data using the private key of the collection device. Integrity verification information can be a hash digest of the asset package's main content, excluding integrity verification information. Authorization rules are preset access control policies, such as accessible requester identifiers, valid usage time ranges, secondary distribution permissions, and decryption limit limits. These components are then serialized and assembled into a structured data packet, and the hash digest of the asset package's main content is calculated as the integrity verification information, forming an encrypted asset package. The encrypted asset package can be stored, transmitted, and distributed independently.

[0071] Among them, Protobuf or CBOR binary serialization methods can be used to serialize the data inside the encrypted asset package. For example, based on this method, the serialized components are assembled into a structured data package.

[0072] Specifically, after data collection is completed or preset fragmentation conditions are met, the zero-trust data collection gateway can encapsulate encrypted data, data lists, scene metadata, device signatures, hash chain roots, authorization rules, watermark parameters, and collection quality indicators into an encrypted asset package. Alternatively, it can encapsulate encrypted data along with scene metadata, device origin proof, integrity digest, and authorization rules into an encrypted asset package. This solution can bind scene keys, device origin proof, hash chains, authorization rules, and multimodal data formats into an encrypted asset package, enabling embodied intelligent data to have registration, retrieval, transaction, and acceptance capabilities without submitting plaintext.

[0073] The encrypted asset package is compatible with common data formats for embodied intelligence, such as MCAP, HDF5, LeRobot dataset structure, or other robot training data formats. For security metadata that cannot be directly contained in the original format, bypass manifest files, index files, or package-level metadata files can be used for recording.

[0074] After receiving the encrypted asset package, the cloud-based authorization and delivery platform can verify the digest, signature, hash chain, and metadata within the package, and generate an asset ID for it. The cloud-based authorization and delivery platform only stores the encrypted data, index, proof, and authorization rules, and does not store the plaintext data. After asset registration is completed, the asset can be retrieved and listed according to dimensions such as scenario, action, device type, data format, collection quality, and authorization scope.

[0075] The preset fragmentation conditions may include at least one of the following: the collection duration reaches a preset duration threshold, the amount of encrypted data in the fragment reaches a preset capacity threshold, the number of multimodal sample frames collected reaches a preset frame number threshold, the collection task ends, the scene identifier changes, a collection stop instruction is received, etc. When any preset fragmentation condition is met or the collection is completed, the zero-trust collection gateway can perform encrypted asset package encapsulation.

[0076] S504. The encrypted asset package is sent to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party through the cloud-based authorization delivery platform, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization certificate to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

[0077] In practice, the requester submits an order, which can be used to request the use of encrypted asset packages corresponding to one or more scenarios. This order can include at least the buyer's identity, asset ID, intended use, usage period, number of uses or duration, whether resale or sublicensing is permitted, usage environment information, budget or price, and compliance commitments. The order and policy engine can then read the authorization policy corresponding to the asset package to determine whether the order meets the requirements of the data owner, data collector, scenario provider, and platform. Furthermore, if the data is owned by a single party, the cloud-based authorization delivery platform can automatically approve or request the data owner's signature confirmation based on preset rules. If the data is jointly owned, the cloud-based authorization delivery platform can request authorization shares from multiple rights holders based on preset threshold rules, such as the data collector, scenario owner, request initiator, platform, or regulator, where t is a positive integer, and authorization is only granted after the signatures of any t entities.

[0078] In this embodiment, the embodied intelligent collection scenario can be used as the minimum key control domain. By using "one scenario, one key", source encryption, encrypted asset package encapsulation, and order policy-driven authorized delivery can be achieved, enabling real-world scenario data to complete trusted circulation without delivering plaintext. The scenario key is then bound to device source proof, hash chain, authorization rules, buyer environment proof, threshold joint authorization, and audit settlement to form an end-to-end zero-trust data infrastructure from edge collection to cloud transaction and then to buyer use.

[0079] In this embodiment, firstly, the scene identifier and its bound scene key can be pre-acquired on the data acquisition side. Source encryption is then performed on the multimodal data generated within the scene, blocking the plaintext leakage path from the data generation stage and preventing the original sensing data from being exposed during transmission and cloud uploading. Secondly, the encrypted asset package encapsulates encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules, enabling security policies to be carried along with the assets. This eliminates the need for separate cloud-based permission configuration maintenance, ensuring that authorization logic, origin evidence, and integrity verification information are not lost during cross-platform transfer of the asset package. Subsequently, the cloud-based authorized delivery platform does not store plaintext. After receiving an order from the requester, it reads the authorization rules built into the asset package to complete the authorization verification. Once the verification is successful, it only outputs the authorization certificate without issuing the plaintext. This decouples data decryption permissions from business access permissions. The authorization certificate constrains the scope of use by the requester, achieving the secure delivery effect of "data not being decrypted, authorization being controllable, and plaintext not being stored on the platform" for embodied intelligent multimodal datasets. This solves the technical defects of existing embodied intelligent datasets, such as plaintext cloud storage, separation of permissions and data, direct distribution of raw data during delivery, data leakage, untraceable source, and coarse-grained permission control.

[0080] Optionally, when the multimodal data corresponding to the encrypted asset package is data on shared ownership by multiple parties, the following steps may also be included: A1. Obtain the multi-party property rights list corresponding to the collection scenario and the preset t-of-n threshold strategy; the t-of-n threshold strategy includes the threshold parameter t and the total number of shares n, where n is the total number of co-owners in this scenario and t is the minimum effective authorization share required to generate the authorization certificate; A2. Divide the scene key bound to the scene identifier into n key shares, and allocate an independent key share to each rights holder; A3. Generate corresponding authorized shares for each rights holder based on their key shares, and distribute them to the corresponding co-owners. The cloud-based authorization delivery platform continuously collects the authorization shares returned by each rights holder. When the number of valid authorization shares collected reaches a preset threshold t, a valid authorization certificate is generated. Only when multiple rights holders who meet the threshold number jointly authorize the decryption of the corresponding encrypted multimodal data can they apply for decryption based on the authorization certificate.

[0081] In practical implementation, in embodied intelligent data collection scenarios involving multi-robot joint collection and shared ownership by multiple parties, multimodal data from the same collection scenario may belong to multiple cooperating rights holders. To prevent a single rights holder from privately decrypting or abusing the data, this embodiment of the application allows for threshold-based secret sharing of the scenario key on the collection party's side. Specifically, the collection party can obtain a list of multiple ownership rights corresponding to the current collection scenario and a preset t-of-n threshold policy. This t-of-n threshold policy can be pre-configured and bound to a scenario identifier; for example, different scenario identifiers can correspond to different t-of-n threshold policies.

[0082] Here, parameter n represents the total number of all co-owners in this scenario, and parameter t represents the minimum number of valid authorized shares required for a legal authorization. Both n and t can be positive integers.

[0083] Subsequently, the data collector can secretly share and split the scenario key corresponding to the current data collection scenario. For example, it can evenly split the key into n independent and unique key shares, so that each co-owner holds a unique key share. The data collector then generates authorization shares for authorization verification based on the key shares of each rights holder, and distributes the corresponding authorization shares to the terminals of each co-owner for storage.

[0084] Furthermore, when a data usage order is subsequently initiated by a demander, the cloud-based authorization delivery platform can continuously aggregate the valid authorization shares returned by each co-owner; the cloud-based authorization delivery platform can only generate a legally valid authorization certificate when the number of valid authorization shares collected reaches a preset threshold t.

[0085] In this embodiment of the application, when the multimodal data corresponding to the encrypted asset package is data jointly owned by multiple parties, the encrypted multimodal data cannot be decrypted by any single rights holder. The threshold condition of joint authorization by multiple co-owners must be met in order to obtain legal decryption permission. This realizes multi-party ownership confirmation, co-management and controllability of data collected by embodied intelligence, and effectively prevents the risk of unilateral disclosure or unauthorized use of shared datasets.

[0086] In this embodiment, for multi-party co-owned data jointly collected by multiple organizations using embodied intelligence, a dedicated t-of-n threshold strategy is configured for the corresponding collection scenario. This strategy clarifies the total number of co-owners and the minimum authorization threshold, accurately adapting to the business characteristics of joint collection and shared ownership of embodied data. Furthermore, by splitting the scenario key into multiple independent key shares and allocating them one-to-one to each rights holder, the original single scenario decryption permission is broken down into multi-party co-managed permissions. This eliminates the risk of unauthorized decryption and transfer of shared data by a single party from the key source. Based on the cloud platform, the authorization shares of each rights holder are aggregated, and a valid authorization certificate is generated only when the preset threshold is met. This strictly constrains the preconditions for data decryption and usage, achieving decentralized management, joint ownership confirmation, and controllable delivery of shared embodied intelligence multimodal data. This effectively avoids the defects of unclear responsibilities, misuse, leakage, and difficulty in tracing the source of shared data, significantly improving the security and compliance of the transfer of jointly collected data assets.

[0087] Optionally, the multimodal data includes at least one fragment of first multimodal data; the step of encrypting the multimodal data generated based on the scene key corresponding to the scene identifier at the source to obtain ciphertext multimodal data may include the following steps: B1. Dynamically generate a unique fragment data key DEK for the first multimodal data of the current fragment; B2. Encrypt the first multimodal data using the segmented data key DEK to obtain the ciphertext multimodal data of the current segment; The above steps, which encapsulate the encrypted multimodal data into an encrypted asset package, can be implemented as follows: The scenario key is used to encrypt and encapsulate the fragmented data key DEK to obtain the DEK ciphertext; The encrypted multimodal data of the current fragment and the DEK ciphertext are encapsulated together into the encrypted asset package; The scenario key is only used to encapsulate the DEK and does not directly encrypt the original multimodal data. Different data shards correspond to independently generated shard data keys (DEKs).

[0088] Specifically, a hierarchical key architecture can be used on the acquisition side to complete the source encryption of fragmented multimodal data. That is, in the embodied intelligent acquisition task, multimodal data can be divided into multiple fragments according to the acquisition duration, data volume or task segment (episode). Each fragment corresponds to a set of first multimodal data. For example, the first multimodal data may include images, point clouds, and IMU sensing data.

[0089] Next, the data acquisition party can dynamically and randomly generate a unique data key DEK for the first multimodal data of the current segment, and then use the DEK corresponding to the segment to encrypt the original first multimodal data in this segment to obtain the ciphertext multimodal data of the current segment.

[0090] Furthermore, the collector then calls the scene key bound to the collection scene to encrypt and encapsulate the sharding data key DEK used in this sharding, generating DEK ciphertext, and encapsulates the ciphertext multimodal data of the current shard and the encapsulated DEK ciphertext together into the ciphertext asset package.

[0091] In this embodiment, the scenario key is only used to encrypt and protect the fragmented data key DEK, and does not directly encrypt the original multimodal data. Each data fragment can independently generate its own fragmented data key DEK, and the DEKs of different fragments are different. When it is necessary to decrypt a fragment of data later, the scenario key must first be used to decrypt the ciphertext of the DEK to obtain the fragmented data key DEK, and then the DEK is used to decrypt the corresponding fragmented ciphertext multimodal data. This method achieves secure isolation between fragments. Even if the DEK of one fragment is leaked, it will not affect the data security of other fragments in the same scenario. At the same time, the scenario key uniformly manages all fragmented DEKs, maintaining scenario-level access control capabilities, and balancing the encryption efficiency and overall security level of large-volume multimodal data.

[0092] In this embodiment, for fragmented embodied intelligent multimodal data, a fragmented data key (DEK) is independently generated for each fragment. The DEK then directly encrypts the original multimodal data of the fragment, adapting to the massive sensor data generated by continuous acquisition and ensuring the efficiency of encryption operations. Furthermore, a scenario key is used to encrypt and encapsulate only the DEK, without participating in the encryption of the original multimodal data. This achieves the separation of business authorization keys and data encryption keys. Different fragments use independent DEKs, achieving fragment-level security isolation. The leakage of a single fragment key will not cause the leakage of all multimodal data in the entire scenario, thus balancing encryption performance and data security in embodied intelligent large-scale fragmented acquisition scenarios.

[0093] Optionally, the following steps may also be included: C1. Obtain the order number, the demander identifier, and the authorization validity period corresponding to the current data collection task as the original parameters for watermark binding; C2. Generate unique traceability watermark parameters that are strongly bound to the order based on the original parameters; C3. The source watermark parameters are secretly embedded into the package-level metadata of the encrypted asset package so that the decrypted and recovered multimodal data comes with an order binding watermark. In the event of data leakage, the corresponding order, user and authorized scenario can be located in reverse based on the source watermark parameters to achieve source tracing and accountability for leakage.

[0094] The data collector can pre-embed order-binding traceability watermark parameters during the encrypted asset package packaging stage. These watermark parameters can be used for subsequent location and accountability in the event of a data leak. Specifically, the data collector can obtain the order number, requester identifier, and authorization validity period corresponding to the current data collection task as the original parameters for watermark binding. Based on these original parameters, a unique traceability watermark parameter is generated, which is bound one-to-one with the order in which the data was used. Then, the traceability watermark parameter is secretly embedded in the package-level metadata area of ​​the encrypted asset package, and the metadata is uploaded to the cloud-based authorization and delivery platform along with the encrypted asset package.

[0095] Furthermore, once the requester completes decryption with valid authorization credentials, the recovered multimodal data will carry the source watermark parameters corresponding to that order. If this multimodal data is leaked, illegally distributed, or resold, the source watermark parameters can be extracted from the leaked data to reverse-engineer the corresponding order number, the requester entity, and the authorization scenario, thereby pinpointing the source of the data leak and completing the tracing of the leak and determination of responsibility.

[0096] Among them, the traceability watermark parameters can be written into the metadata of the encrypted asset package during the collection and packaging stage, without the need to perform additional tagging operations after decryption, thereby avoiding the problem of manual tagging being tampered with or omitted after decryption.

[0097] In this embodiment, during the stage of encapsulating the encrypted asset package by the data collector, a traceability watermark is generated using the order number, the requester identifier, and the authorization validity period as the original parameters. This achieves a strong binding between the watermark and the order for each data usage, ensuring that each authorized delivery has an independent watermark fingerprint. The watermark parameters are then embedded into the package-level metadata of the encrypted asset package, allowing the watermark to circulate along with the encrypted asset package. This eliminates the need for additional watermark addition operations after decryption, avoiding the risk of watermark deletion or tampering after decryption. Furthermore, the decrypted and recovered multimodal data comes with an order-bound watermark. When data leakage occurs, the order, user, and authorization scenario can be traced back based on the watermark, enabling traceability and accountability after the leakage of high-value embodied intelligent multimodal datasets and improving the auditability of dataset circulation.

[0098] Of course, in practice, the traceability watermark parameters can also be secretly embedded in the authorization certificate to be issued, which is not limited here.

[0099] Optionally, the following steps may also be included: D1. Establish a communication connection with the data source device and obtain the device identity credential and sensor operating status parameters of the data source device; D2. Based on the pre-stored trusted device whitelist, the identity credentials of the data source device are authenticated to determine whether the data source device is a legitimate authorized device, and the working status parameters of the data source device are verified to determine whether the sensor is in normal working condition. D3. When the data source device is successfully authenticated and its status is verified, the multimodal data output by the data source device is received and used for subsequent encryption and asset package encapsulation. D4. When the authentication of the data source device fails and / or the device status verification fails, the data output by the data source device shall be discarded and not included in the encrypted asset package.

[0100] The data acquisition party can implement a pre-access authentication and status verification mechanism for the data source devices that are connected to ensure that the multimodal data included in the encrypted asset package all come from trusted and normally functioning sensor devices.

[0101] Specifically, the zero-trust data acquisition gateway can establish communication connections with various data source devices on the robot's end to proactively obtain the device identity credentials and real-time sensor operating status parameters of each data source device.

[0102] Among these, device identity credentials can be understood as unique device identifiers, legitimate device access certificates, etc., specifically authentication information for device authorization. Sensor operating status parameters can include sensor online status, calibration status, sampling frame rate, effective number of point cloud points, sensor signal-to-noise ratio, data offset, etc., which are operational parameters that characterize the sensor's working quality.

[0103] In practice, a locally pre-stored whitelist of trusted devices can be retrieved, and then the identity credentials of the access data source devices can be compared against this whitelist to complete device legitimacy authentication and determine whether the currently accessing device is an authorized trusted device. Simultaneously, in conjunction with preset sensor normal operating thresholds, the acquired sensor operating status parameters are verified for compliance to determine whether the sensor is in normal data acquisition mode. Both the trusted device whitelist and the preset sensor normal operating thresholds can be preset or set by system default.

[0104] Furthermore, if the data source device is successfully authenticated and the sensor status is verified, it can be confirmed that the device is a legitimate and trustworthy data acquisition device and that the acquisition quality meets the standards. In this case, the multimodal data output by the data source device can be received normally and incorporated into the subsequent source encryption, fragmentation processing and encrypted asset package encapsulation process.

[0105] Conversely, if the data source device authentication fails, and / or the sensor's working status is unqualified, it indicates that there are problems such as unauthorized device access, sensor failure, calibration failure, or abnormal data acquisition. In such cases, the multimodal data output by the device can be directly discarded and not included in the encrypted asset package to prevent abnormal or forged data from entering the dataset asset package.

[0106] In this embodiment, a data source device authentication mechanism is added to the data acquisition front end. This mechanism verifies device access through a trusted device whitelist, effectively blocking unauthorized devices from accessing forged data and preventing external malicious data from being injected into the acquisition link. Simultaneously, the sensor's working status and acquisition quality parameters are also verified. This automatically filters out dirty and invalid data caused by sensor malfunctions, calibration anomalies, and low signal-to-noise ratios. Only trusted data that passes the dual verification is included in the encrypted asset package. This ensures the security and acquisition quality of embodied intelligent multimodal data from the source, improving the credibility and usability of the final dataset.

[0107] Please see Figure 6 , Figure 6 This is another flowchart illustrating a data processing method for embodied intelligence data provided in this application embodiment, applied to a cloud-based authorization delivery platform. This data processing method for embodied intelligence data includes: S601. Receive and store a encrypted asset package sent by the collector. The encrypted asset package is obtained by the collector encapsulating encrypted multimodal data. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the collector encrypting the multimodal data generated by the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud-based authorization delivery platform does not store plaintext embodied intelligence data.

[0108] The cloud-based authorization and delivery platform, upon receiving a encrypted asset package, verifies its digest, signature, hash chain, and metadata, and generates an asset ID. The platform only stores the encrypted data, index, proof, and authorization rules; it does not store the plaintext data. After asset registration, the asset can be retrieved and listed based on dimensions such as scenario, action, device type, data format, collection quality, and authorization scope.

[0109] S602. Receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order.

[0110] The requesting party can submit an order requesting the use of one or more encrypted asset packages corresponding to specific scenarios. The order must include at least the buyer's identity, asset ID, intended use, usage period, number of uses or duration, permission for resale or sublicensing, usage environment information, budget or price, and compliance commitments. The order and policy engine reads the corresponding authorization policy for the asset package to determine whether the order meets the requirements of the data owner, data collector, scenario provider, and platform.

[0111] Furthermore, if the data is owned by a single party, the cloud-based authorization and delivery platform can automatically approve or request the data owner's signature confirmation according to the data owner's preset rules; if the data is jointly owned, the cloud-based authorization and delivery platform can request authorization shares from multiple rights holders according to preset threshold rules, such as authorization can only be granted after any t entities among the data collector, scene owner, demand initiator, platform or regulator sign, where t is a positive integer.

[0112] S603. After the verification is passed, the document is not directly issued to the requesting party, but an authorization certificate is output. The authorization certificate is used to restrict the requesting party's access rights to the encrypted asset package.

[0113] After the order policy is approved, the cloud-based authorized delivery platform can generate an order authorization token (AuthToken) and execute any of the following methods or combinations thereof based on the actual delivery mode: key package authorized delivery, proxy re-encryption delivery, trusted use within TEE, threshold joint unblocking, limited number of deliveries or limited period delivery, etc., without any restrictions.

[0114] Among them, key package authorization delivery means that the cloud authorization delivery platform uses the buyer's public key, buyer's UKey public key or buyer's trusted environment public key to re-encapsulate the SceneKey or the DEK encapsulated by the SceneKey, so that only the buyer's authorized environment can decrypt the corresponding key.

[0115] Among them, the agent re-encryption delivery means that the cloud-based authorized delivery platform does not decrypt the original data, but only generates re-encryption parameters based on the authorization of the data rights holder, converting the ciphertext intended for the original rights holder or platform key system into ciphertext intended for the buyer's key system.

[0116] In this context, trusted use within the TEE means that after the cloud-based authorized delivery platform verifies the buyer's remote TEE proof, it releases the authorization key into a verified trusted execution environment, within which the buyer can only perform training, evaluation, or feature extraction.

[0117] Among them, threshold joint unsealing means that for jointly owned data, multiple authorizing parties issue key shares or authorization signatures respectively, and only after the threshold conditions are met will a deliverable key package or re-encryption parameters be generated.

[0118] Among them, the limited number of uses or limited period of delivery means that the authorization token contains the validity period, number of uses, scope of accessible data and revocation mark. The buyer's environment needs to complete token verification or online / offline counting each time it is used.

[0119] Furthermore, the buyer (demand party) obtains data usage capabilities constrained by the order policy, rather than unlimited ownership of plaintext. In the key package delivery mode, the buyer can only decrypt the corresponding scenario key within their private key, UKey, or trusted environment; in the TEE mode, plaintext data only appears briefly in the trusted execution environment; in the proxy re-encryption mode, the cloud-based authorized delivery platform can complete targeted delivery without handling the plaintext.

[0120] The cloud-based authorization and delivery platform records audit logs for order authorization, key release, re-encryption, download, TEE verification, usage count, revocation, and revenue distribution. For assets requiring accountability, data watermarks associated with the order, buyer identity, or authorization token can be embedded during the encrypted encapsulation or authorization delivery stage. In the event of a leak or dispute, the cloud-based authorization and delivery platform can locate the authorization path based on the watermark and audit logs.

[0121] In practice, the release of scenario keys, DEK key packages, or re-encryption parameters can be controlled based on order and authorization policies, rather than simply providing plaintext downloads. Furthermore, the key authorization, delivery, and usage processes are recorded, enabling data assets to be traceable and auditable. Order-policy-driven key authorization and re-encryption delivery processes allow data usage rights to be technically constrained according to purpose, duration, frequency, environment, and resale rules.

[0122] In this application embodiment, TEE remote proof verification can be used to verify the buyer's training or decryption environment before releasing the key. Threshold cryptography, multi-party signatures, or multi-party key shares can be used to achieve joint authorization of jointly owned data. Proxy re-encryption can also be used to enable the platform to complete targeted encrypted delivery without decrypting the original data. Order-related watermarks can also be embedded in the data asset package or authorized delivery results to achieve accountability for leakage.

[0123] Furthermore, this application's embodiments employ a hierarchical key structure, key version management, and revocation list to improve the efficiency of key rotation and revocation in large-scale scenarios. Audit summaries can also be written to a blockchain, trusted log, or tamper-proof log system to enhance the ability to provide evidence in disputes.

[0124] In this embodiment, the cloud-based authorization delivery platform stores only the encrypted asset package uploaded by the data collector, without storing any plaintext embodied intelligence data. This completely eliminates the risk of plaintext data leakage and unauthorized distribution from the cloud storage perspective. Furthermore, scene metadata, device origin proof, integrity verification information, and authorization rules are uniformly encapsulated within the encrypted asset package. This achieves integrated binding of data, traceability evidence, integrity credentials, and authorization policies with the asset, avoiding issues such as uncontrolled permissions, lack of traceability, and data tampering caused by the separation of cloud permission configuration and data ontology. Subsequently, after receiving the data usage order from the requester, the cloud-based authorization delivery platform can complete the authorization policy verification based on the authorization rules built into the asset package. Once the verification is successful, only the authorization credentials that restrict usage permissions are issued, and plaintext data is never directly output. This transforms the data access mode into a zero-trust delivery mode with credential authorization, controllable decryption, and no plaintext data being sent to the cloud. This achieves full lifecycle controllability, traceability, and non-illegal leakage of embodied intelligence multimodal data, significantly improving the security of sharing and transferring high-value training datasets for robots.

[0125] Optionally, the following steps may also be included: E1. Receive the remote proof report of the trusted use environment uploaded by the requester. The remote proof report is generated by the trusted execution environment of the requester based on the hardware root of trust. E2. Parse the remote certification report to obtain environmental compliance information, which includes at least one of the following: environmental metric value, firmware version, running whitelist, and unique trusted hardware identifier; E3. Compare and verify the environmental compliance information with the authorization conditions corresponding to the encrypted asset package to determine whether the current environment used by the requester is a trusted and compliant environment. E4. The process of releasing the scene key is only allowed if the environment meets the preset authorization conditions and the authorization certificate has been verified to be valid. Otherwise, if the environment verification fails, the release of the scene key will be prohibited and decryption will be refused.

[0126] Hardware-level trusted verification of the decryption environment of the requester can be performed on the cloud-based authorization and delivery platform to achieve controlled release of the scenario key. Specifically, when the requester applies for a scenario key to decrypt the encrypted asset package using authorization credentials, the requester's internal trusted execution environment can generate a remote proof report based on the hardware root of trust and upload the remote proof report to the cloud-based authorization and delivery platform.

[0127] Subsequently, the cloud-based authorization and delivery platform can receive the remote certification report and parse its contents to extract environmental compliance information. This environmental compliance information includes at least one of the following: environmental metrics, firmware version, operating whitelist, and unique trusted hardware identifier. The cloud-based authorization and delivery platform will then compare and verify the extracted environmental compliance information with the authorization conditions built into the encrypted asset package to determine whether the requester's current decryption environment meets the preset security requirements.

[0128] Furthermore, the cloud-based authorization delivery platform will only allow the release of the scenario key if both the usage environment verification meets the preset authorization conditions and the authorization certificate verification is valid. Conversely, if the remote proof report verification result indicates that the usage environment is non-compliant, the release of the scenario key will be directly prohibited, and the requesting party will be refused the ability to decrypt and use the encrypted asset package.

[0129] The encrypted asset package can be pre-packaged with preset authorization conditions, which may include trusted environment benchmark metrics, allowed firmware version ranges, a whitelist of executable programs, and a unique trusted identifier for legitimate hardware. The cloud-based authorization delivery platform compares the environment compliance information parsed from the remote verification report with these preset authorization conditions item by item to determine whether the requester's trusted execution environment meets the security baseline for decrypting the encrypted asset package. This mechanism can bind the key release to the trusted hardware environment, preventing the key from being distributed to untrusted environments such as virtual machines or tampered systems for decryption, avoiding the migration, copying, and reverse engineering of encrypted data, and ensuring that the key can only be used within a trusted environment.

[0130] In this embodiment, a remote verification report can be generated using the hardware root of trust of the requester. This remote verification report can objectively reflect the underlying operating status of the terminal and is difficult to be forged by software. It enables hardware-level trusted verification of the decryption environment. The cloud-based authorization delivery platform parses the report to obtain environmental compliance information such as environmental metrics, firmware version, and running whitelist, and compares it with the preset authorization conditions corresponding to the encrypted asset package to achieve fine-grained environmental access judgment. Furthermore, a dual verification mechanism of authorization certificate legality and trusted environment is adopted. The scenario key is only allowed to be released when both verifications pass, preventing the key from being sent to an illegal environment for decryption. This prevents high-value embodied intelligent multimodal data from being migrated, copied, and leaked, achieving data usability but not migration, and further improving the end-to-end zero-trust delivery system.

[0131] For example, consider a scenario of targeted encrypted delivery for an exclusive data collection order. Client A can request the collection of dexterity motion data from the data collector, requiring both the collector and the cloud-based authorized delivery platform to refrain from storing plaintext data. The cloud-based authorized delivery platform creates a SceneID-A and binds the public key UKey provided by Client A to this scene. A zero-trust data collection gateway connects to a robotic arm, camera, and force sensor at the data collection site, using SceneKey-A to encrypt the image, trajectory, and force data of this scene at the source, generating a encrypted asset package. After data collection, the cloud-based authorized delivery platform only receives the encrypted asset package and proof of origin. Upon order acceptance, the cloud-based authorized delivery platform encapsulates the SceneKey-A or the corresponding DEK key package and assigns it to Client A's UKey, preventing the collector and the cloud-based authorized delivery platform from reading the plaintext data.

[0132] To illustrate further, consider the scenario of subsequent reuse transactions of shared ownership datasets. Collector B, scenario owner C, and cloud-based authorization delivery platform D can agree that the action data of a certain production line workstation is shared ownership data. During collection, the zero-trust collection gateway uses SceneKey-B to encrypt the data source and splits SceneKey-B into three-party key shares. The rule is that any two parties agree to authorize a third party to use it. Third-party client E places an order to purchase this data for model evaluation. The cloud-based authorization delivery platform requests authorization from B, C, and D based on the order. After B and C issue authorization shares, the cloud-based authorization delivery platform generates re-encryption parameters, converting the encrypted asset package into an encrypted or key package usable by E, and automatically generates revenue sharing records according to preset revenue rules. This threshold-based joint authorization mechanism for shared ownership data allows the collector, scenario owner, client, platform, or regulator to jointly control subsequent reuse, preventing any single entity from unilaterally selling or unsealing data.

[0133] To illustrate further, consider the scenario where training is available but not visible within the TEE. Client F wants to train a model using high-risk operational data, but the data provider requests that plaintext be avoided. The cloud-based authorized delivery platform requires F to provide a remote verification report of the training environment, confirming that it is running the agreed-upon training image, code version, and trusted hardware execution environment. After successful verification, the cloud-based authorized delivery platform can release the order authorization token and scenario key to the TEE environment. The encrypted asset package is then decrypted within the TEE and used for training. After training, only model parameters, evaluation metrics, or intermediate results permitted by the policy are output. The platform records the environment verification and audit summary used.

[0134] In this embodiment, an end-to-end closed loop combining edge-based source encryption with cloud-based authorized transactions is implemented, balancing seamless access at the data collection site, secure cloud circulation, and use by the buyer's own infrastructure. Data is encrypted and encapsulated from the source, ensuring that the data collector, platform provider, and intermediate processing stages do not handle plaintext by default, reducing the risks of copying, retention, and resale, thereby mitigating the risk of plaintext leakage. Furthermore, a unique key for each scenario controls usage rights at the level of scenario, task, purpose, duration, frequency, and resale rules, preventing unauthorized use due to coarse-grained account or file permissions, thus improving data authorization accuracy. Additionally, device identity, collection scenario, data fragment, hash chain, and signature proof are all bound together, enabling the requesting party to verify whether the data originates from the agreed-upon device and scenario, enhancing source credibility.

[0135] Furthermore, the platform can complete encrypted asset registration, authorized delivery, and audit settlement without storing plaintext or training large models, making it suitable for highly sensitive scenarios such as industry, energy, warehousing, and special operations to support the circulation of highly sensitive data. Through threshold authorization or multi-party signatures, subsequent transactions of shared ownership data require joint authorization from the agreed-upon parties, and can automatically generate separate ledger records to support shared ownership reuse transactions. Of course, the platform does not need to be the custodian of plaintext data, but rather acts as a coordinator for encrypted assets, authorization strategies, and audit settlement, reducing the consequences of data leaks caused by platform attacks or internal unauthorized actions, thus lowering the platform's trust burden.

[0136] Furthermore, by using hash chains, watermarks, authorization tokens, and audit logs, the source, recipient, scope of authorization, and leakage path of data can be traced, thereby improving the ability to provide evidence in disputes. The edge gateway is adaptable to existing robotic arms, sensors, and remote-operated devices without altering the main control processes at the data collection site. The cloud platform handles transactions, delivery, and auditing, facilitating deployment within existing data collection site orders and enhancing industry applicability.

[0137] Consistent with the above embodiments, please refer to Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. As shown in the figure, the electronic device includes a processor, a memory, a communication interface, and one or more programs. The one or more programs are stored in the memory and configured to be executed by the processor.

[0138] In this embodiment of the application, when the electronic device includes a data acquisition device, the above-mentioned program includes instructions for performing the following steps: Obtain the scene identifier of the collection scene corresponding to the embodied intelligence data and the scene key bound to the scene identifier; Based on the scene key, the multimodal data generated by the scene corresponding to the scene identifier is encrypted at the source to obtain ciphertext multimodal data; The encrypted multimodal data is encapsulated into an encrypted asset package, which encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules; The encrypted asset package is sent to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party through the cloud-based authorization delivery platform, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization document to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

[0139] or, In this embodiment of the application, when the electronic device includes a cloud-based authorized delivery platform, the above-mentioned program includes instructions for performing the following steps: The system receives and stores encrypted asset packages sent by the data acquisition party. These encrypted asset packages are obtained by the data acquisition party by encapsulating encrypted multimodal data. Each encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin verification, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the data acquisition party through source encryption of multimodal data generated in the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud-based authorization delivery platform does not store plaintext embodied intelligence data. Receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order; After successful verification, instead of issuing an official document to the requesting party, an authorization certificate is output. The authorization certificate is used to restrict the requesting party's access rights to the encrypted asset package.

[0140] Figure 8 This is a schematic diagram of a data processing device for embodied intelligence data applied to a data acquisition party, according to an embodiment of this application. The data processing device 800 for embodied intelligence data applied to a data acquisition party includes: an acquisition unit 810, an encryption unit 820, an encapsulation unit 830, and a transmission unit 840, wherein... The acquisition unit 810 is used to acquire the scene identifier of the collection scene corresponding to the embodied intelligent data and the scene key bound to the scene identifier; The encryption unit 820 is used to perform source encryption on the multimodal data generated by the scene corresponding to the scene identifier based on the scene key, so as to obtain ciphertext multimodal data. The encapsulation unit 830 is used to encapsulate the encrypted multimodal data into an encrypted asset package. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The transmission unit 840 is used to send the encrypted asset package to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization document to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

[0141] It is understood that the functions of each program module of the data processing device for embodied intelligent data applied to the data acquisition party in this embodiment can be based on the above. Figure 5 The specific implementation of the method in the method embodiment shown can be referred to the relevant description of the above method embodiment. It can also implement other methods on the data acquisition side, which will not be elaborated here.

[0142] Figure 9 This is a schematic diagram of the structure of a data processing device oriented towards embodied intelligence data applied to a cloud-based authorization delivery platform, provided in an embodiment of this application. The data processing device 900 oriented towards embodied intelligence data applied to a cloud-based authorization delivery platform includes: a receiving unit 910, a verification unit 920, and an output unit 930, wherein... The receiving unit 910 is used to receive and store the encrypted asset package sent by the collector. The encrypted asset package is obtained by the collector encapsulating encrypted multimodal data. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the collector encrypting the multimodal data generated by the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud authorization delivery platform does not store the embodied intelligence data plaintext. The verification unit 920 is used to receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order. The output unit 930 is used to output an authorization certificate instead of directly issuing an invention document to the requesting party after the verification is passed. The authorization certificate is used to restrict the requesting party's access rights to the encrypted asset package.

[0143] It is understood that the functions of each program module of the data processing device for embodied intelligence data applied to the cloud-based authorization delivery platform in this embodiment can be based on the above. Figure 6 The specific implementation of the method in the method embodiment shown can be referred to the relevant description of the above method embodiment. It can also implement other methods on the data acquisition side, which will not be elaborated here.

[0144] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the methods described in the above method embodiments, wherein the computer includes an electronic device.

[0145] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments. The computer program product may be a software installation package, and the computer may include an electronic device.

[0146] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0147] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0149] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0150] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0151] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0152] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0153] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A data processing method for embodied intelligence data, characterized in that, Applied to the data collection party, the method includes: Obtain the scene identifier of the collection scene corresponding to the embodied intelligence data and the scene key bound to the scene identifier; Based on the scene key, the multimodal data generated by the scene corresponding to the scene identifier is encrypted at the source to obtain ciphertext multimodal data; The encrypted multimodal data is encapsulated into an encrypted asset package, which encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules; The encrypted asset package is sent to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party through the cloud-based authorization delivery platform, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization document to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

2. The method according to claim 1, characterized in that, When the multimodal data corresponding to the encrypted asset package is data on shared ownership by multiple parties, the method further includes: Obtain the multi-party property rights list corresponding to the collection scenario and the preset t-of-n threshold strategy; the t-of-n threshold strategy includes a threshold parameter t and a total number of shares n, where n is the total number of co-owners in this scenario and t is the minimum effective authorization share required to generate the authorization certificate; The scene key bound to the scene identifier is split into n key shares, and each rights holder is allocated an independent key share. Based on the key shares of each rights holder, a corresponding authorized share is generated and distributed to the corresponding co-ownership rights holder; The cloud-based authorization delivery platform continuously collects the authorization shares returned by each rights holder. When the number of valid authorization shares collected reaches a preset threshold t, a valid authorization certificate is generated. Only when multiple rights holders who meet the threshold number jointly authorize the decryption of the corresponding encrypted multimodal data can they apply for decryption based on the authorization certificate.

3. The method according to claim 1 or 2, characterized in that, The multimodal data includes at least one fragment of first multimodal data; the source encryption of the multimodal data generated based on the scene key and the scene corresponding to the scene identifier to obtain ciphertext multimodal data includes: Dynamically generate a unique fragment data key DEK for the first multimodal data of the current fragment; The first multimodal data is encrypted using the segmented data key DEK to obtain the ciphertext multimodal data of the current segment; The step of encapsulating the encrypted multimodal data into an encrypted asset package includes: The scenario key is used to encrypt and encapsulate the fragmented data key DEK to obtain the DEK ciphertext; The encrypted multimodal data of the current fragment and the DEK ciphertext are encapsulated together into the encrypted asset package; The scenario key is only used to encapsulate the DEK and does not directly encrypt the original multimodal data. Different data shards correspond to independently generated shard data keys (DEKs).

4. The method according to claim 1 or 2, characterized in that, The method further includes: The order number, the demander identifier, and the authorization validity period corresponding to the current data collection task are obtained as the original parameters for watermark binding. Generate unique traceability watermark parameters that are strongly bound to the order based on the original parameters; The source watermark parameters are secretly embedded into the package-level metadata of the encrypted asset package, so that the decrypted and recovered multimodal data comes with an order binding watermark. In the event of data leakage, the corresponding order, user and authorized scenario can be located in reverse based on the source watermark parameters, so as to trace the source and hold accountable the leakage behavior.

5. The method according to claim 1 or 2, characterized in that, The method further includes: Establish a communication connection with the data source device to obtain the device identity credentials and sensor operating status parameters of the data source device; The identity credentials of the data source device are authenticated based on a pre-stored trusted device whitelist to determine whether the data source device is a legitimate authorized device, and the working status parameters of the data source device are verified to determine whether the sensor is in normal working condition. When the data source device is successfully authenticated and its status is verified, the multimodal data output by the data source device is received and used for subsequent encryption and asset package encapsulation. When the data source device fails authentication and / or the device status verification fails, the data output by the data source device is discarded and not included in the encrypted asset package.

6. A data processing method for embodied intelligence data, characterized in that, The method, applied to a cloud-based authorization delivery platform, includes: The system receives and stores encrypted asset packages sent by the data acquisition party. These encrypted asset packages are obtained by the data acquisition party by encapsulating encrypted multimodal data. Each encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin verification, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the data acquisition party through source encryption of multimodal data generated in the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud-based authorization delivery platform does not store plaintext embodied intelligence data. Receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order; After successful verification, instead of issuing an official document to the requesting party, an authorization certificate is output. The authorization certificate is used to restrict the requesting party's access rights to the encrypted asset package.

7. The method according to claim 6, characterized in that, The method further includes: Receive the remote proof report of the trusted use environment uploaded by the requester, wherein the remote proof report is generated by the trusted execution environment of the requester based on the hardware root of trust; Parse the remote certification report to obtain environmental compliance information, which includes at least one of the following: environmental metrics, firmware version, running whitelist, and unique trusted hardware identifier. The environmental compliance information is compared and verified with the authorization conditions corresponding to the encrypted asset package to determine whether the current environment used by the requester is a trusted and compliant environment. The scenario key release process is only allowed if the environment meets the preset authorization conditions and the authorization certificate has been verified to be valid. Otherwise, if the environment verification fails, the scenario key release is prohibited and decryption is refused.

8. A data processing device for embodied intelligent data, characterized in that, Applied to the data acquisition side, the device includes: an acquisition unit, an encryption unit, an encapsulation unit, and a transmission unit, wherein, The acquisition unit is used to acquire the scene identifier of the collection scene corresponding to the embodied intelligent data and the scene key bound to the scene identifier; The encryption unit is used to perform source encryption on the multimodal data generated by the scene corresponding to the scene identifier based on the scene key, so as to obtain ciphertext multimodal data. The encapsulation unit is used to encapsulate the encrypted multimodal data into an encrypted asset package. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The transmission unit is used to send the encrypted asset package to the cloud-based authorization delivery platform. The cloud-based authorization delivery platform does not store the embodied intelligent data plaintext. It receives data usage orders issued by the demand party, reads the authorization rules corresponding to the encrypted asset package, performs authorization policy verification on the data usage order, and outputs an authorization certificate instead of directly issuing an authorization document to the demand party after the verification is passed. The authorization certificate is used to restrict the demand party's usage rights of the encrypted asset package.

9. A data processing device for embodied intelligent data, characterized in that, The device, used in a cloud-based authorization delivery platform, includes: a receiving unit, a verification unit, and an output unit, wherein... The receiving unit is used to receive and store the encrypted asset package sent by the collector. The encrypted asset package is obtained by the collector encapsulating encrypted multimodal data. The encrypted asset package encapsulates at least the following data: the encrypted multimodal data, scene metadata, device origin proof, integrity verification information, and authorization rules. The encrypted multimodal data is obtained by the collector encrypting the multimodal data generated by the scene corresponding to the scene identifier based on the scene key. The scene key is bound to the scene identifier. The cloud authorization delivery platform does not store the embodied intelligence data plaintext. The verification unit is used to receive a data usage order issued by the demander, read the authorization rules corresponding to the encrypted asset package, and perform authorization policy verification on the data usage order. The output unit is configured to, after successful verification, not directly issue an invention document to the requesting party but instead output an authorization certificate, which is used to restrict the requesting party's access rights to the encrypted asset package.

10. An electronic device, characterized in that, It includes a processor and a memory, the memory being used to store one or more programs and configured to be executed by the processor, the programs including instructions for performing the steps in the data processing method for embodied intelligence data as described in any one of claims 1-7.